• HamsterRage@lemmy.ca
    link
    fedilink
    arrow-up
    19
    ·
    3 months ago

    Back in the 70’s and 80’s there were “Travesty Generators”. You pushed some text into them and they developed linguistic rules based on probabilities determined by the text. Then you could have them generate brand new text randomly created by applying the linguistic rules developed from the source text.

    Surprisingly, they would generate “brand new” words that weren’t in the original text, but were real words. And the output matched stylistically to the input text. So you put in Shakespeare and you got out something that sounded like Shakespeare. You get the idea.

    I built one and tried running some TS Eliot through it, because stuff is, IMHO, close to gibberish to begin with. The results were disappointing. Basically because it couldn’t get any more gibberishy that the source.

    I strongly suspect that the same would happen with Trump’s gibberish. There used to be a bunch of Travesty Generators online, and you could probably try one out to see.

  • Possibly linux
    link
    fedilink
    English
    arrow-up
    6
    ·
    3 months ago

    You don’t even need to train it. Just use a system prompt.

    Just don’t ask it for medical advise

    • otter@lemmy.dbzer0.com
      link
      fedilink
      English
      arrow-up
      5
      arrow-down
      10
      ·
      3 months ago

      Downvotes without representation are, as always, bullshit, so let’s clarify: it doesn’t take many brain cells to imagine how easily such a thing would (will?) be used to credit that walking cum rot for whatever the"AI" spits out…

      No is the only answer with any foresight.